Anti-AI-slop: a 4 staged skill to detect and remove any AI writing slop

Anti-AI-slop is a skill for coding agents such as Claude Code and Codex that removes AI-writing tells from a draft and then checks the result before returning it. The skill targets sentence patterns, words and phrases, rhythm and formatting, and issues that only a second reviewer catches. It operates through four gates: first, it removes invisible characters like zero-width spaces and hidden tag characters; second, a Python linter scores the draft out of 100 against detectable patterns, requiring a score of 90 with no serious findings; third, a second agent reviews the original and rewrite to catch made-up facts, made-up connections, changed meaning, dropped caveats, and synonym cycling; and fourth, an optional model review using another model such as Gemini 3.8 Flash through LiteLLM. If a check fails, the draft goes back for another pass up to two times, and unresolved issues are listed for the user. The skill can be installed for Claude Code or other agents, and the checks require Python 3.8 or newer. The optional model review also requires LiteLLM. Users can run the checks directly via scripts like sanitize.py, slopcheck.py, and grade.py, and slopcheck.py exits with 0 on a pass and 1 on a fail, making it suitable for pre-commit hooks or CI. Rules are stored in rules.json and can be extended.